Two kinds of definites in numeral classifier languages

Numeral classifier languages distinguish definite noun phrases licensed by uniqueness from those licensed by familiarity. Unique definites are expressed by bare nouns or null pronouns, while familiar definites are expressed by indexicals such as demonstrative descriptions or overt pronouns. This generalization parallels the observation by Schwarz (2009) that German distinguishes unique versus familiar or anaphoric definiteness in its article system. The difference between the two kinds of definites can be reduced to the presence of a semantic index in the case of familiar definites. As familiar definites occur in most E-type contexts, including donkey anaphora, and uniqueness definites are not possible in these contexts, these facts provide support to dynamic analyses of E-type anaphora and pose problems for uniqueness-based approaches, such as the theory of Elbourne (2013).